Pulse Separation Using Independent Component Analysis

Abstract

This paper intends to demonstrate pulse separation using independent component analysis (ICA). Independent component analysis is related to the blind source separation (BSS) problem when independent components or the original signals are unknown or blind. BSS is explained by the cocktail party problem where groups of people have separate conversations at a cocktail party, and an individual is selectively extracting a conversation from a mix of conversations. The blind source separation problem is like deinterleaving signals. When a radar receiver intercepts multiple signals, the signals go through interleaving, where the waveforms mix with each other. The process of separating the mixed signals is called deinterleaving. In the radar system, finding the pulse descriptor word (PDW) is essential for deinterleaving because the parameters from PDW will help reconstruct the received radar signals. The radar will process the parameters from the PDW to construct pulse trains to help identify all the source signals that were detected by the radar receiver. Independent component analysis is a method to identify all the source signals received by the receiver.