Handling Side-Information for Data Compression of Radar Pulse Trains

Previously a method has been proposed for a high-performance compression method designed expressly for compressing intercepted radar pulse trains for the purpose of locating the transmitter. The method relies on gating pulses, putting them into a pulse matrix and then using the singular value decomposition (SVD) to compress the signal data. The pulse gating step, however, generated a sequence of varying integers as side information and the previous study did not provide a means for efficiently coding this side information; although it did given an empirically generated estimate of the entropy of this side information. This paper reformulates the pulse gating method to generate a different form of side information that is much easier to compress. It is shown that this new side information can be simply and efficiently coded at a rate that compares favorably with the empirically-estimated entropy of the original method. It is shown that this new pulse gating method requires the solution of an integer linear programming problem and several standard methods are first considered. It is shown that the large number of constraints in the original formulation can be significantly reduced by replacing the constraint set by its convex hull; simple rules for identifying the convex hull are given. However, even with these reductions the execution time for these methods can be prohibitive at very large pulse counts; furthermore, these methods exhibited numerical precision and convergence problems as the number of pulses increased. Therefore, an efficient non-standard method for solving this integer optimization problem is developed by exploiting characteristics of the objective function. This method solves the pulse gating problem with short execution times that grow negligibly with increasing pulse counts.