Maximizing the Convergence Rate of the EM Algorithm for the Multiple Broad-Band Signal Estimation and Detection.

摘要

The following results were obtained under the contract: numerical acceleration of convergence for the EM algorithm; a characterization of the rank-n ambiguity; a characterization of the phase delay ambiguity; determination of the minimal number of known and unknown signals for sensor location observability; a large class of cyclic regression algorithms for multiple signal direction finding; array sensor localization and calibration by cyclic regression; cyclic regression for the weighted subspace fitting for direction finding: cyclic regression for harmonic retrieval from colored noise; a statistical approach to training multilayer perceptrons.