Topic: Sensor Arrays and Information Theoretic Fusion
By: Alex Ihler (MIT), John Fisher (MIT), Alan Willsky (MIT)
Large arrays of sensors with overlapping regions of influence and multiple sensing modalities pose new problems in data fusion. Complex relationships between the sensor observations, including reflections and nonlinear effects, make it difficult to determine such quantities as number, location, and characteristics of sources, and even the sensor array configuration. This is related to such problems as ICA (independent component analysis), sensor fusion, and more standard array processing problems like finding direction of arrival. Our approach uses nonparametric estimates of mutual information to attempt to determine relationships between sensors, and frame the extraction of the original sources as an ICA-like problem.
Questions and comments to the webmaster. This page is copyrighted by the Massachusetts Institute of Technology.